1. Field of the Invention
The present invention relates to an imaging apparatus, and more particularly to an imaging apparatus having a pixel-defect detecting function.
2. Description of the Background Art
Imaging apparatuses such as video cameras have been widely used. Electronic still cameras for mainly picking up and recording still pictures have recently been popular as digital cameras. Further, video movie cameras for mainly recording moving pictures have come to incorporate a still-picture pickup/record function. A long-exposure technique is known as a technique for enabling photography even in a low illumination, without using auxiliary lighting equipment such as a flash. This long-exposure technique is mainly used to pick up a still picture, and lengthen the exposure period by lengthening a charge-accumulating period in an imaging element.
In an imaging element such as a CCD, there is a dark output resulting from, for example, a so-called dark current. The dark output is superposed upon an image signal, which may degrade the image quality. A pixel of a high dark output level is called a defective pixel. Whether a pixel is defective is determined, for example, in the following manner. The dark output is estimated by employing a predetermined standard exposure period (for example, in NTSC, 1/60 sec. or a period obtained by adding thereto a predetermined margin, 1/15 sec. in the case of e.g. a four-time margin). If the dark output level is high, the pixel is considered defective.
A method for complementing information using the output from pixels adjacent to a defective pixel instead of the defective pixel is widely employed. In this specification, complement processing will be referred to as “pixel defect compensation”.
Further, pixel defects depend upon temperature and may occur with time. In light of this, a technique for improving the point that the estimation of defective pixels is insufficient if it is executed only before shipping products from a factory has been proposed (see Japanese Patent Application KOKAI Publication No. 06-038113). In this technique, detection and compensation of a defective pixel is executed as follows. Immediately after the turn on of the power supply, the iris is closed to shield the light-receiving surface from light. After that, the dark output from the imaging element is estimated before using the camera, thereby detecting, as a defective pixel, a pixel having a dark output with a predetermined detection level or more. Defect compensation is then executed on each detected defective pixel.
However, if a defect detecting function is provided in a camera, it is necessary to newly register detected defect data in a non-volatile memory, such as an EEPROM, in which initial defect data obtained in a factory is pre-registered. In this case, if, for example, battery is exchanged while defect data is being written to the EEPROM, there is a fear that the initial defect data is lost. The initial defect data is very reliable data acquired by repeatedly analyzing image information at a predetermined temperature. Therefore, if the initial defect data is lost, a sufficient compensation cannot be executed, thereby degrading the image quality.
Moreover, even in the case of employing the technique for acquiring defective-pixel information using the output from the imaging element in a light-shielded state with the iris closed, it is possible, in a digital camera with a single-reflex optical viewfinder, that reverse incident light through the optical viewfinder, for example, may enter the CCD as unwanted light and cause a detection error. In this case, for example, a central portion of the CCD is irradiated with a light spot of an indefinite outline, and the level of the entire central portion is significantly increased. As a result, all the pixels at the central portion are erroneously detected as defective. At this time, if defect compensation is executed on the basis of defective-address data, the resultant image quality is more degraded than in a case where no defect compensation is executed.
Such a problem may be caused by light guided through an unexpected route (even if, for example, the iris serving as imaging/light-shielding means is out of order and cannot be completely opened, normal imaging can be executed when the shutter normally operates), as well as reverse incident light through the optical viewfinder. In addition, a similar problem may occur if a number of defects concentrically occur for some other system error.
In the meantime, the inventors have found from their recent research that a defect as should be called a “flicker-type defect” exists, too. This flicker-type defect means a defective pixel that behaves in the following manner. While an imaging process (charge accumulation and data reading) is being repeated under identical imaging conditions including the temperature and exposure (accumulation) period, etc., a certain pixel sometimes behaves as a white defect (excessively large dark charge), and sometimes outputs a normal signal, i.e., it behaves as if a white defect (excessively large dark charge) flickers.
The cause of the flicker-type defect has not yet been theoretically determined. However, the following information concerning the phenomenon has been acquired.
This flicker phenomenon seems to have a certain probabilistic element. Further, it does not have a predetermined cyclic property nor depend upon the number of reading times. Furthermore, there is a pixel that flickers in an imaging process repeated in a relatively short period (this will hereinafter figuratively be referred to as a “pixel of a short cyclic property”, although the flicker phenomenon is not cyclic as aforementioned), and a pixel of a relatively long cyclic property. A possibility is indicated that post defects having been considered to be a destructive (non-repairable) phenomenon caused by natural radiation or cosmic radiation would have contained flicker-type defects of an extremely long cyclic property. On the other hand, it has been clarified that a great number of short-period flicker type post defects exist.
Since various flicker-type defects exist, it is obvious that the conventional defect registration method and immediately preceding detection method are both useless. The defect registration method is a method of using a defect address registered in a factory. The immediately preceding detection method is a method for using a defect address newly acquired by a defect detection executed, for example, when power is turned on. Further, just a combination of these methods is not sufficient. It is indispensable to additionally register and use a newly detected defect address.
Such an additional registration system, however, involves another problem, i.e., a problem of additionally registering data that is not suitable for additional registration. Data that is not suitable for additional registration includes a normal pixel erroneously detected during defect detection for some reason (hereinafter referred to as an “erroneously detected defect”), or a defect temporarily occurring for a reason different from the aforementioned “flicker phenomenon” (hereinafter referred to as a “temporal defect”), etc.                (1) Although an erroneously detected defect has to be dealt with by the detection system, it can always exist, unnoticed by the detection system because of its technical limit.        (2) As an example of a temporal defect, there is a case where a great number of defects occur due to a temporal but extreme increase in the temperature of the apparatus at a high ambient temperature.        
In this case, since defects are occurring at the present stage, it is necessary to execute defect compensation during imaging. However, they are not defective pixels if the apparatus is in a normal state of use (in a room-temperature environment).
Once an erroneously detected defect or temporal defect is registered, it is still considered a defect even after the state is restored to a normal state of use, whereby its signal data is canceled. This is disadvantageous since the original image quality cannot be obtained. From another viewpoint, the registration of such a pixel wastes the memory capacity, which may cause, at worst, a serious problem in which a real post defect cannot be registered. A real post defect means a post defect that exists even at room temperature.
However, a temporal defect is a currently occurring defect. Therefore, it may be subjected to defect compensation for imaging. Further, erroneously detected defects may include a real defect as well as erroneously detected data. If it is not subjected to defect compensation, it will be actualized. In light of this, they must be subjected to defect compensation as temporal defects. At this time, the erroneously detected data is also compensated. However, this does not significantly degrade the image quality. Further, since erroneously detected defects are only temporal defects, no problem will be caused by registration if they are not registered.
As described above, where detected defect data is considered to contain erroneously detected data or temporal defects, it is an object suitable for defect compensation. However, the detected defect data is not an object suitable for additional registration. In other words, the reliability of the detected defect data is insufficient as a to-be-additionally-registered object. The prior art does not consider this point.